Learning Obstacle Representations for Neural Motion Planning

Strudel, Robin; García, Ricardo Navarro; Carpentier, Justin; Laumond, Jean–Paul; Laptev, Ivan; Schmid, Cordelia

doi:10.48550/arxiv.2008.11174

Cited by 3 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many neural motion-planning methods try to encode the workspace with deep neural networks. In [95], Strudel et al proposed to learn a function that encodes the obstacles and the goal configuration as a vector. A PointNetlike network is used to encode point cloud in their method.…”

Section: Motion Planning With Policy-based Rl Methodsmentioning

confidence: 99%

A survey of learning‐based robot motion planning

Wang

Zhang

et al. 2021

IET Cyber-Syst and Robotics

View full text Add to dashboard Cite

A fundamental task in robotics is to plan collision-free motions among a set of obstacles. Recently, learning-based motion-planning methods have shown significant advantages in solving different planning problems in high-dimensional spaces and complex environments. This article serves as a survey of various different learning-based methods that have been applied to robot motion-planning problems, including supervised, unsupervised learning, and reinforcement learning. These learning-based methods either rely on a human-crafted reward function for specific tasks or learn from successful planning experiences. The classical definition and learning-related definition of motion-planning problem are provided in this article. Different learning-based motion-planning algorithms are introduced, and the combination of classical motion-planning and learning techniques is discussed in detail.This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

show abstract

Section: Motion Planning With Policy-based Rl Methodsmentioning

confidence: 99%

A survey of learning‐based robot motion planning

Wang

Zhang

et al. 2021

IET Cyber-Syst and Robotics

View full text Add to dashboard Cite

show abstract

“…By leveraging classical components and scaling up learning, we are able to learn models that generalize to novel objects. Learning for motion planning has been used to reduce the runtime of motion planning algorithms: [15], [16], [43]. Strudel et al [43] learn obstacles representations for motion planning, while Ichter et al [15], [16] use learning to bias sampling of states for motion planners.…”

Section: Related Workmentioning

confidence: 99%

“…Learning for motion planning has been used to reduce the runtime of motion planning algorithms: [15], [16], [43]. Strudel et al [43] learn obstacles representations for motion planning, while Ichter et al [15], [16] use learning to bias sampling of states for motion planners. Our use of learning is similarly motivated, but we learn to predict low-dimensional strategies (that can be decoded into full motion plans) for constrained motion planning problems from visual input.…”

Section: Related Workmentioning

confidence: 99%

Public Financing in Higher Education Institutions: Issues and Challenges

Gupta

2018

Shiksha Shastra Saurabh

View full text Add to dashboard Cite

This paper is an attempt to analyze the present pattern of funding higher education in Nepal and to discuss the desirability and feasibility of various alternative methods of funding the same. Before federal system, higher education (HE) in Nepal was a state funded but currently the paradigm has been shifted to the province government besides the funding of central university, which has not been well-defined. Higher education itself benefits not only society at large, but also individuals, and as it attracts relatively more privileged sections of the society, there is a rational for shifting the financial burden to the social domain rather to government sector alone. The main objective is to review the current policies, practices of financing in promoting access, quality, relevancy, equity and prosperity. On the basis of secondary sources of data this paper focused the reports published by University Grants Commission, Nepal (UGC). Some other relevant literature, reports, articles, books, policy documents were critically analyzed and reviewed to draw the findings and conclusion. Some statistical methods were used for interpretation. Resource constraints, equity, financing HE mostly from the general tax revenue may not be a desirable policy in the long run. Accordingly some of the alternative policy choices are discussed for financing HE from the public exchequer, student loans, student fees, and the role of the community investment. Among the available alternatives, it is argued that a discriminatory pricing mechanism and subsidy policies to community campuses of different layers of government would be relatively more efficient and equitable mechanism of financing as community campuses are providing the major access, equity, quality as well and reducing the financial burden of government with the cost recovery modality. It is argued that such institutions should be the priority of government for promoting equitable quality access to HE with nominal costs of users.

show abstract

“…These methods first encode the environmental point cloud data and then exploit neural networks to fit the expert’s demonstration trajectories, but these methods depend on the quality of the dataset and are subject to error accumulation when the neural networks are forward-propagated to generate samples. Reinforcement learning approaches treat the motion planning problem as a Markov process [ 25 ], where the intelligence learns the planning strategy through continuous trial and error, yet the manipulation skills learned by the robot in the simulation environment are difficult to deploy on real robots. Deep reinforcement learning-based planning algorithms have lots of model parameters and it is difficult to deploy on robotic arms.…”

Section: Introductionmentioning

confidence: 99%

An Efficient Motion Planning Method with a Lazy Demonstration Graph for Repetitive Pick-and-Place

Zuo

et al. 2022

Biomimetics

View full text Add to dashboard Cite

Robotic systems frequently need to plan consecutive similar manipulation in some scenarios (e.g., pick-and-place tasks), leading to similar motion plans. Moreover, the workspace of a robot changes with the difference in operation actions, which affects subsequent tasks. Therefore, it is significant to reuse information from previous solutions for new motion planning instances to adapt to workplace changes. This paper proposes the Lazy Demonstration Graph (LDG) planner, a novel motion planner that exploits successful and high-quality planning cases as prior knowledge. In addition, a Gaussian Mixture Model (GMM) is established by learning the distribution of samples in the planning cases. Through the trained GMM, more samples are placed in a promising location related to the planning tasks for achieving the purpose of adaptive sampling. This adaptive sampling strategy is combined with the Lazy Probabilistic Roadmap (LazyPRM) algorithm; in the subsequent planning tasks, this paper uses the multi-query property of a road map to solve motion planning problems without planning from scratch. The lazy collision detection of the LazyPRM algorithm helps overcome changes in the workplace by searching candidate paths. Our method also improves the quality and success rate of the path planning of LazyPRM. Compared with other state-of-the-art motion planning algorithms, our method achieved better performance in the planning time and path quality. In the repetitive motion planning experiment of the manipulator for pick-and-place tasks, we designed two different experimental scenarios in the simulation environment. The physical experiments are also carried out in AUBO−i5 robot arm. Accordingly, the experimental results verified our method’s validity and robustness.

show abstract

Learning Obstacle Representations for Neural Motion Planning

Cited by 3 publications

References 0 publications

A survey of learning‐based robot motion planning

A survey of learning‐based robot motion planning

Public Financing in Higher Education Institutions: Issues and Challenges

An Efficient Motion Planning Method with a Lazy Demonstration Graph for Repetitive Pick-and-Place

Contact Info

Product

Resources

About